OUTLET'S EXTRA =================
    I N S T R U C T I O N S     
================================

Bug Fixes for Betados and G+DOS 
        by Miles Kinloch       
-------------------------------
I have found and fixed a number
of bugs in Betados - some fairly
minor, others more serious -
plus some bugs which are also
present in G+DOS. The program
which follows in this issue,
"BUGTESTER". demonstrates the
effects of each bug, showing
clearly what happens before and
after they are fixed.

A further two programs, GDOSFIX
and BETAFIX in ADDENDERS in this
issue, will make the required
corrections to the respective
DOS.

Note that to squeeze the extra
code into the +D RAM has meant
the removal of the Boot message
and also some of the disc-copy
messages.

[Also note that OUTLET has been
using the bytes in the Boot
message as a simple test between
the DOSs. This test will be void
and may cause corrupt- ions once
you have installed the "Fixes"
Ed.]

  DOS BUG FIXES & ENHANCEMENTS
  ----------------------------

The program "BETAFIX" cures the
following bugs in Betados and
also adds a couple of new
features. Where the bug also
affects G+DOS, it can be cured
by the program "GDOSFIX".

BUGS
----
1. OPEN # command. If you omit
the IN or OUT at the end, the
DOS will not accept the syntax
if there is another statement
following in the same line,
i.e. if a colon is the next
character after the command.
(This bug also affects G+DOS,
and is fixed by "GDOSFIX").

2. FORMAT command. There are two
bugs here:

 a) The special form FORMAT d1,n
 etc,ignores the drive specified
 and just formats the last drive
 used.

 b) The form FORMAT d* will not
 pass syntax.

3. Command codes. These are
perhaps the most serious of all
the Betados bugs, as they affect
many commercial programs which
call the DOS from machine-code.
Once the command codes are
fixed, the great majority of
such software will then run with
Betados without any problems.

A few examples are the +D conv-
erted version of Tasword 128,
Hackers Workbench(SD Software),
Profile (Glensoft), Disc-Kit (in
OUTLET #76) and several of
Betterbytes' utilities including
the disc-copier one. Six codes
altogether are faulty which are:

 a) Codes 52 and 53. The problem
 here is that the alternate HL
 register pair is corrupted,
 resulting in a crash after
 returning to BASIC from a USR
 call which involves these.

 b) Code 56. The routine called
 by RST 8 DB 56 has not been
 changed from the one in the +D
 ROM which G+DOS uses. This is
 fine with Betados too, EXCEPT
 if used on a disk that has been
 formatted for over 80 files.
 Such a disk would invariably be
 corrupted as a result. (Code 56
 is used to close a file from
 machine-code, i.e. create an
 entry for it in the directory.)

 c) Code 67. This is the code
 used to CAT a disk, and is fat-
 ally flawed. Any attempt to use
 it will almost always crash the
 Spectrum and possibly corrupt
 the DOS at the same time.

 d) Codes 68 and 69. The equiv-
 alents, in fact, of SAVE @ and
 LOAD @. These two always return
 signalling a DOS error, regard-
 less of whether there has actu-
 ally been one or not. See also
 No.4, below.

4. LOAD @/SAVE @ commands (and
command code equivalents). If
there has been an error during a
LOAD @ or SAVE @, e.g. a write-
protected disk etc.,this results
in a wrong value being left in
one of the DOS system variables.
Consequently, if you then type
RUN to re-boot the DOS, the pro-
cess will stop prematurely with
an 'END of file' message. Also,
if you try to make a snapshot,
it will likely be corrupted and
could possibly corrupt other
files on the disk too. (Typing
CLEAR # resets the variable, but
this is hardly the ideal way
round it, i.e. if you had any
streams open at the time or if
the error had come about through
the command codes and you could
not exit to BASIC.)

5. SAVE @ command. There is a
problem with the routine which
spins the drive for about a
second to ensure the motor is up
to speed before writing the
sector. It doesn't take account
of the drive specified, and will
just spin the last drive used.
This is fine if you have a disk
in both drives, but if the last
drive used no longer has the
disk in it, its motor will just
spin indefinitely until one is
inserted.

6. Use of a capital 'D' in SAVE
and ERASE commands. Contrary to
appearances, the auto-CAT which
occurred with G+DOS when using
the capital 'D' has not in fact
been fully disabled in Betados.
The CAT actually still takes
place, it's just that you can't
see it as its stream is altered
to make it CAT into the BASIC
workspace instead of the screen.
In these circumstances no checks
are made on the amount of
workspace memory available etc,
and this can sometimes cause
problems, especially if an INPUT
has been used previously as this
also uses the workspace.

7. MOVE command. The form MOVE #
TO d1"name" (not documented in
the +D manual) has two serious
bugs. The first is that the
directory entry created for the
new file, although it will be
there, won't be visible in the
catalogue.

The second problem is that, when
using this form of the MOVE
command, if the filename has
already been used and you type
'n' at the 'Overwrite' prompt,
using CLEAR # or CLOSE #* after-
wards will lock up the Spectrum.
This is because some pointers
won't have been reset, and the
CLEAR # or CLOSE #* will get
'confused' about how much memory
to reclaim(This bug also affects
G+DOS and is cured by"GDOSFIX").

8. Betados FN calls, e.g. DEF FN
s(x)=USR 8 etc.Potentially there
could be a problem with the 128K
Editor if these were typed as
direct commands, particularly
the ones concerned with opentype
files, e.g. FN e, FN l and FN p.

The 128K editor is fussy about
the value of the IX register
pair (or rather, to be more
precise, the contents of (IX+1)
if non-zero) when it's in the
process of editing a line. The
latter three functions use IX
themselves so this should really
be set correctly afterwards for
100% safety.

For further details about this
IX problem,see the article 'Plus
D Alters'in Format Magazine,
March 1988 (Vol. 1, No. 8), as
this was something which also
affected the previous DOS (i.e.
the version before G+DOS 2a).

9. The tape SAVE command. If you
use a null filename or one
longer than 10 characters, you
get "Nonsense in BASIC" instead
of "Invalid file name" when a +D
is attached. A minor bug perhaps
but one which is very easy to
sort(This bug also affects G+DOS
and is cured by "GDOSFIX").

10. Printing to Random-access
files. There are two bugs here:

 a) The Betados manual states if
 you INPUT from files opened RND
 any text or control characters
 before the input variable are
 ignored,eg, INPUT #4;"text";a$.
 However, this is only true when
 you are using 48K mode. In 128
 BASIC such text will still be
 written to the file regardless,
 just as with files opened OUT.

b) Using colours in a PRINT
 command which sends data to the
 file can cause an error.
 Embedded control codes are OK,
 but if you use the PAPER, INK
 etc. tokens in the PRINT state-
 ment,eg, PRINT #4;INK 5;"text",
 this can give a "Nonsense in
 BASIC" report.

ENHANCEMENTS
------------
1. A New Command is added to
both BETADOS and G+DOS: You can
now use LOAD dn (or SAVE dn) on
its own which gives the opportu-
nity to change the current drive
without actually having to use
the drive to do so. This can be
very handy before LOAD Pn, for
instance.

2. SAVE...TO. This command has
been improved when copying
between different drives. The +D
system suffers from a short-
coming in that it only keeps
note of the track register(drive
head position) for one drive.

This can sometimes lead to prob-
lems, especially when copying
long files such as 128K snap-
shots which can't be copied in a
single 'pass'.

The system gets confused about
where the drive head is, and if
the file being copied has tracks
near the end of a disk, it can
sometimes fail altogether,
stopping with an error report.
BETAFIX patches the SAVE...TO
routines so that a separate
record is kept for the track
registers of both drives,result-
ing not only in much greater
reliability but also an increase
in speed, typically about 15%.

In practice, this problem is
less serious with G+DOS than
with Betados, due to the limit-
ations G+DOS has in the first
place, i.e. its inability to
copy snapshots or opentype
files, where these effects tend
to be worst.

THE FOLLOWING TWO ENHANCEMENTS
APPLY TO G+DOS ONLY, AS THEY ARE
ALREADY INCLUDED IN BETADOS...

3. SAVE @ and LOAD @. A new
patch has been added to these,
similar to the one in Betados,
which spins the drive for about
a second before reading/writing
the sector.For people who suffer
from this symptom,these commands
will now be just as reliable as
their Betados equivalents.

4. Saving to a write-protected
disk.Previously the SAVE command
would first check the directory
etc., and if a file of the same
name already existed, print the
"Overwrite?" prompt etc. before
realising the disk was protect-
ed. This wasted a bit of time.
GDOS-FIX now makes the DOS check
the write-protect status first
when saving, so people will be
informed of this right away.

COMPATIBILITY
-------------
GDOSFIX is fully compatible with
all G+DOS patches and mods
except "THINGi", (OUTLET #77),
which uses all the spare +D RAM.


Finally: USING BDM TO CREATE A
         BETADOS SYSTEM FILE

Only ever use an unmodified
G+DOS 2a for this purpose, i.e.
one that has been made using the
CONFIG program as supplied on
the original "system tape" and
has had no subsequent patches
etc. added since, including the
ones in GDOSFIX. All necessary
bug fixes for Betados will be
taken care of by BETAFIX itself.

[Miles Kinloch mentions that he
found the "ESD Disassembler"
indispensible for tracking down
the various bugs detailed above,
and recommends it for anyone
wanting to look into the +D
ROM/RAM. - Ed.]